Figma’s Design Agent Is Open Beta—Paid Access Still Has Boundaries

Figma’s Design Agent is available on paid Figma plans during open beta. It appears in Figma Design’s left navigation, edits native canvas layers and does not consume AI credits during the beta.

That plan-level entitlement does not remove every boundary. The relevant seat, file access, team AI settings, design-system governance and separate controls for shared libraries and Figma Make still determine what a user can change or publish.

Access matrix: plan eligibility and seat context

Figma’s current Design Agent page lists access on all paid plans, placement in Figma Design’s left navigation and zero AI-credit consumption during open beta. The free Starter plan is outside that paid-plan statement.

• Professional, Organization and Enterprise: These paid plans fall within the current open-beta eligibility statement.
• Full seats: The original rollout explicitly supported Full seats on those three paid plans, making this the clearest documented seat configuration for work in team files.
• Collab and Dev seats: The original rollout limited their use to drafts. Because the current product page does not publish a replacement seat-by-seat matrix, check availability in the intended team file before relying on broader access.
• Education and Government: These were excluded from the initial rollout. The current “all paid plans” wording does not expressly resolve their product-specific status.

The apparent conflict comes from two rollout stages rather than two descriptions of the same moment. Figma’s May 20 launch post documented a gradual beta with a waitlist, Full-seat eligibility on Professional, Organization and Enterprise, draft-only use for Collab and Dev seats, and exclusions for Starter, Education and Government. The newer product page supersedes the waitlist and plan-level availability language, but leaves the detailed seat boundaries unstated.

Run a bounded first task

1. Open the Figma Design file owned by the team where the work will remain. Access demonstrated in a personal draft may not reflect the settings of that workspace.
2. Open the agent from the left navigation. If it is absent, check the file’s workspace, your seat, file access and whether AI features are enabled there.
3. Select the relevant frame or layer. Duplicate it when the task could make broad changes, preserving an immediate comparison point.
4. Describe both the result and the permitted scope. Identify the copy, layout, component state, typography or imagery that may change instead of requesting an unrestricted file-wide revision.
5. Select the design library the result should use and identify any components, variables or tokens that must remain authoritative.
6. Review the resulting layer hierarchy, component instances, variables, modes, constraints, auto layout, content and interaction states before keeping the revision.

Duplicating a frame is a review safeguard, not a permission boundary. It does not replace file access controls, version history or approval from the people responsible for the design system.

What the agent can change—and what still needs review

The agent can generate screens and alternative structures, revise copy and imagery, update typography, switch component states and repeat edits across multiple frames. Its output remains editable Figma content, so individual layers can be inspected and overridden rather than accepted as a flattened asset.

Design-system context improves relevance but does not guarantee compliance. The documented workflow starts with frequently and recently used components, while users can choose a library and reference particular tokens, variables and components. Review whether instances remain attached, the correct variable modes are preserved, approved components were selected and generated content is appropriate for the file’s audience.

The publication boundary is explicit: the agent can build with existing components, styles and variables, but it cannot publish new assets to a shared library by itself. An authorized maintainer must review the changes and perform the library-publication step.

Design Agent, Make and MCP have different jobs

• Design Agent: Works inside Figma Design and creates or edits native design layers using canvas and design-system context.
• Figma Make: Generates code layers for functional prototypes and web apps. Work can move between Design and Make when a team needs to clarify visual intent separately from executable behavior.
• Figma MCP server: Bridges Figma and an external AI client or development environment. Figma’s MCP documentation explains that the server passes structured design context and a code starting point; the external assistant, not the server, generates the final code.

The practical map is: Design Agent changes design layers; Make creates a functional code-layer prototype; MCP supplies Figma context to an external agent workflow. The native agent’s beta credit policy therefore should not be treated as a statement about Make usage, MCP eligibility or publishing rights.

Review design-system, privacy and publishing controls

• AI availability: Confirm that AI features are enabled for the applicable team, workspace or organization. This setting is separate from per-person access to paid AI credits.
• Content training: Check the governing team setting before submitting confidential material. Figma’s AI administration guidance documents training as enabled by default for Starter and Professional teams, disabled for Organization and Enterprise, and separate from access to AI features.
• Edit scope: For broad requests, work from a bounded selection or duplicate and inspect bindings in the layer tree rather than judging only the rendered frame.
• Library publication: Apply the normal maintainer review to agent-edited components and documentation. Editable canvas output is not automatically a released shared-library update.
• Make publication: Treat Make as a separate deployment surface. Figma’s Make publishing instructions specify that publishing makes an app available on the web, while Organization and Enterprise plans can restrict it to an internal audience; password protection and unpublishing provide additional controls.

Open beta removes the Design Agent’s immediate credit charge, not the surrounding governance. A paid plan establishes product eligibility; seat context, file access, AI administration, human review and the relevant publication mechanism determine whether an editable proposal can become a shared asset or live app.

Also read:

• Figma Just Dropped Its Design Agent — And It’s Sitting Right Inside Your File
• Odyssey 2 Pro: Pioneering Real-Time Interactive Video Generation with Open API Access
• The Future of Coding Has No Editor — Just a Prompt and an Agent